Image Processing And 3D Surface Temperature Field Reconstruction Of Thermal Image

In nature,objects whose temperature is higher than absolute zero will continuously emit infrared radiation energy to the environment.By detecting this radiation,the infrared thermal camera outputs image in color space that can be recognized by human eyes to realize the temperature detection.However,due to its low pixel resolution and the human eyes are more used to processing visible light information,the infrared thermal image is not very intuitive.When using infrared thermal images for detection,it may be difficult to distinguish object characteristic according to the temperature distribution,resulting in difficulties for positioning different temperature areas.To solve this problem,based on current study of stereo reconstruction,a 3D temperature reconstruction method fusing visible light information and temperature information is proposed.According to the registration of visible light camera and infrared thermal camera,the proposed method fuses the temperature information on the visible light reconstruction to realize the 3D temperature reconstruction.The proposed method first estimates the camera poses and depth maps using structure from motion and multi view stereo algorithm,which is further optimized by algorithm combining shading and geometric data terms.Then,through a calibration method that can calibrate both cameras,the intrinsic and extrinsic parameters of visible camera and infrared thermal imager are obtained.However,because the ge ometric reconstruction is completed by a monocular system,the scale of the coordinate system between the reconstruction and calibration process is not unified.Therefore,the calibration information cannot be directly used to complete the registration.To unify the scale,according to the assumption that infrared thermal images and visible images have similar the depth distribution,the scale estimation is realized through an algorithm based on mutual information.With the scale being estimated,combining the registration and visible reconstruction results,the 3D dense reconstruction combining visible light information and temperature information is finally realized.